Robotic Episodic Memories with RAG

Supervisor: Ilaria Tiddi (i.tiddi@vu.nl)

Description

In this project, the goal is to work on developing a framework for knowledge-based explainable robots in long-horizon interactive tasks. From experiential robot data, the student will populate robot episodic memories in the format of a time-indexed knowledge graph, and will integrate this memory into a natural language interactive agent (e.g. RAG- based LLM) so that humans can demand explanations about robot experiences. Durning the whole project development, the student will count on our expertise on applied ontology and knowledge graphs for robotics, and we will provide the student with a ‘ready to use’ framework for storing memories from robot data. (Project in collaboration with Barcelona, Spain)